Dear SPM-community,
I hope to get your input on a reviewer's concern with the FlexFactorial module in SPM.
A reviewer noted that the Flexible Factorial Design in SPM (and FSL for that matter) the same denominator (mean sum of square for the residuals) is adopted for all F-stats including main effects and interaction. And that the same testing strategy is propagated to those t-stats in this tool as well.
This would lead to the following problems:
1.
*in a two-way mixed ANOVA (1 between, 1 within factor), the between main effect is wrong (which I knew).
2.
However,
*The reviewer claims that in a two-way within ANOVA, the F-stat for main effect of both within-subject factors is also wrong (which I did not knew). Similarly, for a two-way within ANOVA the t-tests of any within-subject factor could be incorrect, and even for a one-way within subject ANOVA the t-tests would be wrong when the contrast weights do not add up to 0.
Can someone explain point 2 in more detail of why this would happen in the current implementation of SPM?
Or point me to an appropriate reference (unfortunately the reviewer does not provide any references to problem 2)?
Many thanks for your help!
Best wishes,
Anna
|